Note: Descriptions are shown in the official language in which they were submitted.
CA 02615412 2007-11-22
1
D'YNAMIC BANDWIDTH ALLOCATION FOR MULTIPLE ACCESS
COMMUNICATION USING SESSION QUEUES
BACKGROUND OF THE INVENTION
The increasing use of wireless telephones and personal computers has led to
a corresponding demand for advanced telecommunication services that were once
thought to only be used in specialized applications. In the 1980's, wireless
voice
communication beoame widely available through the cellular telephone network.
Such services were at first typically considered to be the exclusive province
of the
businessman because of expected high subscriber costs. The same was also true
for
access to remotely distributed computer networks, whereby until very recently,
only
business people and large institutions could afford the necessary computers
and
wireline access equipment. As a result of the widespread availability of both
technologies, the general population now increasingly wishes to not only have
access to networks such as the Internet and private intranets, but also to
access such
networks in a wireless fashion as well. This is particularly of concern for
the users
of portable computers, laptop computers, hand-held personal digital assistants
and
the like who would prefer to access such networks without being tethered to a
telephone line.
There still is no widely available satisfactory solution for providing low
cost,
high speed access to the Internet, private intranets, and other networks using
the
existing wireless infrastructure. This situation is most likely an artifact of
several
unfortunate circumstances. For one, the typical manner of providing high speed
data
service in the business environment over the wireline network is not readily
adaptable to the voice grade service available in most homes or offices. Such
standard high speed data services also do not lend themselves well to
efficient
transmission over standard cellular wireless handsets. Furthermore, the
existing
cellular network was originally designed only to deliver voice services. As a
result,
the emphasis in present day digital wireless communication protocols and
CA 02615412 2007-11-22
-2-
modulation schemes lies with voice, al ough certain schemes do provide some
measure of asytnmetrical behavior for e accommodation of data transrxtission.
For example, ttxe data rate on an IS-95 forward traffic channel can be
adjusted in
increments from 1.2 kilobits per seco (kbps) up to 9.6 kbps for so-catled Rate
Set I and in for increments from 1.8 k s up to 14.4 Icbps for Rate Set 2. On
the
reverse link tra.ffic channel, however, data rate is fL:ed at 4.8 kbps.
The design of such existing syst ms ttterefore typically provides a radio
channel which can accommodate maxiia utn data rates oztly in the range of 14.4
kilobits per second (lcbps) at best iu the orvvard direotion. Such a low
ci,ata rate
channel does not lend itself directly to 'tting data at a rate of56.6 kbps
which
is now commonly available using inexp 've wire ]iiue modetas, not to me.ntion
even higher rates such as the 1281cbps hich is available with lntegrat.ed
Services
Digital Network (ISDN) type quipm Data rates at these lewe]s are rapidly
beconaing the mitimurn acceptable rate for activities such as browsi,n.g web
pages.
Other types of data networks using Itigh speed building blocks such as Digital
Subscriber Line (xDSL) service are conniug into use in the United States.
However, their costs have only been tly reduced to the point wherc they at~e
attractive to the residential custouzer .
Although such networlCs vcfere lai wn at the time that cellular systems were
oiiginally deployed, for the most part th is no provision foz providing higher
speed ISDN- or xDSL-grade data servic s over cellular network topologios.
Unfortunately, in wireless env:ironments accees to channels by multiple
subscxibers is expensive and tb,ere is co etition for them. Whether the
multiple
access is providod by the traditional Frec. y D'avision Mnltiple Access (FDMA)
using analog modulation om a group of 'o carriers, or by newer digital
modulation schemes the permit sharing a radio canrier using Time Division
Multiple Access (TDIVLA.) or Code Di'vis n Multiple Access (CDMA), the nature
of the radio specstiun2 is that it is a m.edi that expected to be shared. This
is
quite dissizailar to the traditional en ' ent for data trar,smi,ssion, in.
wlrioh the
wire line medituu is relatively iuncxpensi to obtain, and is tYrerefore not
typieally
intended to be shared.
Other considerations are the char teristics of the data itself. F or exanaple,
consider that access to web pages in gen al is burst-oriented, with as~-
mmetrical
data rate transmission requirements. In p'cular, the user of a remote client
CA 02615412 2007-11-22
3
computer first specifies the address of a web page to a browser program. The
browser program then sends this web page address data, which is typically 100
bytes
or less in length, over the network to a server computer. The server computer
then
responds with the content of the requested web page, which may include
anywhere
from 10 kilobytes to several megabytes of text, image, audio, or even video
data.
The user then may spend at least several seconds or even several minutes
reading the
content of the page before requesting that another page be downloaded.
Therefore,
the required forward channel data rates, that is, from the base station to the
subscriber, are typically many times greater than the required reverse channel
data
rates.
In an office environment, the nature of most employees' computer work
habits is typically to check a few web pages and then to do something else for
extended period of time, such as to access locally stored data or to even stop
using
the computer altogether. Therefore, even though such users may expect to
remain
connected to the Internet or private intranet continuously during an entire
day, the
actual overall nature of the need to support a required data transfer activity
to and
from a particular subscriber unit is actually quite sporadic.
Furthermore, prior art wireless communication systems provide a continuous
bandwidth to individual subscribers. That is, in such networks, during a
communication session the bandwidth available at all times is constant and has
been
designed, as noted above, primarily for voice grade use.
SUMMARY OF THE INVENTION
Prior art methodologies for transmission of data over wireless networks
suffer numerous problems. As noted above, the bandwidth available for a single
subscriber unit channel is typically fixed in size. However, data
communications
tend to be bursty in nature, often requiring a need for large amounts of
bandwidth at
certain times, while requiring very little amounts, or even none, at other
times.
These wide swings in bandwidth requirements can occur very close together in
time.
For example, when browsing a web site using HyperText Transfer Protocol
(HTTP), the user typically selects pages by selecting or clicking a single
link to a
page causing the client computer to send a small page request packet to the
web
server. The request packet in the receive link direction requires very little
CA 02615412 2007-11-22
4
bandwidth. In response to the request, the server,typically delivers one or
more web
pages ranging in size from 10 to 100 kilobits (kB) or more to the client in
the
forward link direction. To receive the pages, the bandwidth requirements are
much
greater than to request the pages. The optimum bandwidth needed to acceptably
receive the pages is rarely realized due to the inefficiency of the present
wireless
protocols that only offer maximum data rates of about 9600 bps under optimal
conditions. This results in the server having to hold back some of the
requested data
until the network can "catch up" with the data delivery and also results in
fiustrated
users having slow response and page loading times. In essence, the bandwidth
to
send a request is more than is needed, and the bandwidth to receive the pages
is not
enough to deliver the data at acceptable rates.
Another problern with prior art systems is that the difference between the
time which the page request message leaves the wireless network and becomes
wirebound, and the time when the pages of requested data enter the wireless
portion
of the data communications session is often quite long. This time-from-request
to
time-of-receipt delay is a function of how congested the network and the
server are.
The present invention is based in part on the observation that bandwidth is
being wasted during periods of time when waiting for data from the wireline
network. Prior art wireless communications systems maintain the constant
availability of the full bandwidth of the 9600 bps wireless connection for
that entire
data communication session, even though the wireless client may be waiting for
retarn pages. This bandwidth which is effectively unused is therefore wasted
because there is no way to allocate the channel resources in use for this data
conununication session to another session needing more bandwidth. That is, if
other
concurrent wireless data conununications sessions are taking place for other
subscriber units, these concurrent sessions have no way in the prior art
systems to
take advantage of any unused bandwidth allocated to the client merely waiting
for
return pages, as in this example.
The present invention provides high speed data and voice service over
standard wireless connections via an unique integration of protocols and
existing
cellular signaling, such as is available with Code Division Multiple Access
(CDMA)
type systems. The invention achieves high data rates through more efficient
allocation of access to the CDMA channels.
CA 02615412 2007-11-22
Specifically, the invention provides a scheme for determining an efficient
allocation of N fixed rate data channels amongst M users. The invention
addresses
the problem of how to allocate these channels in the most effective manner
between
users competing for channel use. For example, when more users exist than
5 channels, the invention determines a set of probabilities for which users
will require
channel access at which times, and assigns channel resources accordingly. The
invention can also dynamically take away or deallocate channels (i.e.,
bandwidth)
from idle subscribers and provide or allocate these freed-up channels to
subscribers
requiring this bandwidth.
Channel resources are allocated according to a buffer monitoring scheme
provided on forward and reverse links between a base station and multiple
subscnber units. Data buffers are maintained for each connection between a
base
station and a subscriber unit. Each buffer is monitored over time for
threshold levels
of data to be transmitted in that buffer. In essence, the thresholds measure
the
"fullness" of buffers over time for each respective subscn'ber unit are
monitored.
For each buffer, a probability is calculated that indicates how often that a
specific
buffer for a specific subscriber will need to transmit data and how much data
will be
transmitted. This probability takes into account the arrival rates of data
into the
buffer, as well as which thresholds within the buffer are exceeded, as well as
which
resources in the form of channels are already allocated to the subscnber unit.
Based
on this probability, channel resources for data transmission can be either
allocated or
deallocated to subscriber units depending upon a forecasted need.
BRIEF DESCRIPTION OF THE DRAWINGS
The foregoing and other objects, features and advantages of the invention
will be apparent from the following more particular description of preferred
embodiments of the invention, as illustrated in the acconipanying drawings in
which
like referencg characters refer to the same parts throughout the different
views. The
drawings are not necessarily to scale, emphasis instead being placed upon
illustrating the principles of the invention.
Fig. 1 is a block diagram of an example wireless communication system
making use of a bandwidth management scheme according to the invention.
CA 02615412 2007-11-22
6
Fig. 2 is a diagram showing how channels are assigned within a given radio
frequency (RF) channel.
Fig. 3 is a diagram illustrating the protocol layers of a wireless
communication system.
Fig. 4 illustrates the structure of session queues and data buffers used in
the
base station.
Fig. 5 is a buffer level diagram.
Fig. 6 is a buffer level diagram when resources are being added.
Fig. 7 is a buffer level diagram when resources are being taken away.
DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT
Turning attention now to the drawings more particularly, Fig. 1 is a block
diagram of a system 100 for providing high speed data service over a wireless
connection by seamlessly integrating a wired digital data protocol such as,
for
example, Transmission Control Protocol / Internet Protocol (TCP/IP) with a
digitally
modulated wireless service such as Code Division Multiple Access (CDMA).
The system 100 consists of two different types of components, including
subscriber units 101-1, 101-2, ..., 101-n (collectively subscribers 101) as
well as one
or more base stations 104 to provide the functions necessary in order to
achieve the
desired implementation of the invention. The subscriber units 101 provide
wireless
data and/or voice services and can connect devices such as, for example,
laptop
computers, portable computers, personal digital assistants (PDAs) or the like
through base station 104 to a network 105 which can be a Public Switched
Telephone Network (PSTN), a packet switched computer network, or other data
network such as the Internet or a private intranet. The base station 104 may
communicate with the network 105 over any number of different efficient
communication protocols such as primary rate ISDN, or other LAPD based
protocols such as IS-634 or V5.2, or even TCP/IP if network 105 is an Ethernet
network such as the Internet. The subscriber units 101 may be mobile in nature
and
may travel from one location to another while communicating with base station
104.
Fig. 1 illustrates one base station 104 and three mobile subscriber units 101
by way of example only and for ease of description of the invention. The
invention
CA 02615412 2007-11-22
7
is applicable to systems in which there are typically many more subscriber
units 101
communicating with one or more base stations 104.
It is also to be understood by those skilled in the art that Fig. 1 may -be a
standard cellular type communication system such as a CDMA, TDMA, GSM or
other system in which the radio channels are assigned to carry between the
base
stations 104 and subscriber units 101. This invention, however, applies more
particularly to non-voice transmissions, and preferably to digital data
transmissions
of varying bandwidths. Thus, in a preferred embodiment, Fig. 1 is a CDMA-like
system, using code division multiplexing principles for the air interface.
However, it
is also to be understood that the invention is not limited to using
standardized
CDMA protocols such as IS-95, or the newer emerging CDMA protocol referred to
as IS-95B. The invention is also applicable to other multiple access
techniques.
In order to provide data and voice connnunications between the subscriber
units 101 and base station 104, wireless transmission of data over a limited
number
of radio channel resources is provided via forward communication channels 110
which carry information from the base station 104 to the subscriber units 101,
and
reverse communication channels 111 which carry information from the subscriber
units 101 to the base station 104. The invention provides dynamic bandwidth
management of these limited channel resources on an as needed basis for each
subscn'ber unit 101. It should also be understood that data signals travel
bidirectionally across the CDMA radio channels 110 and 111, i.e., data signals
originating at the subscriber units 101 are coupled to the network 105, and
data
signals received from the network 105 are coupled to the subscriber units 101.
Fig. 2 provides an example of how dynamic allocation of radio bandwidth
may take place in system 100. First a typical transceiver within a subscriber
unit
101 or the base station 104 can be tuned on command to any 1.25 MegaHertz
(MIiz)
channel within a much larger bandwidth, such as up to 30 MHz in the case of
the
radio spectrum allocated to cellular telephony. This bandwidth is typically
made
available in the range of from 800 to 900 MHz in the United States. For PCS
type
wireless systems, a 5 or 10 MHz bandwidth is typically allocated in the range
from
about 1.8 to 2.0 GigaHertz (GHz). In addition, there are typically two
matching
bands active simultaneously, separated by a guard band, such as 80 MHz; the
two
CA 02615412 2007-11-22
8
matching bands form a forward and reverse full duplex link between the base
station
104 and the subscnlber units 101.
Within the subscriber unit 101 and the base station 170, transmission
processors (i.e., transceivers) are capable of being tuned at any given point
in time
to a given 1.25 MHz radio frequency channel. It is generally understood that
such
1.25 MHz radio frequency carrier provides, at best, a total equivalent of
about a 500
to 600 kbps maximum data rate transmission speed within acceptable bit error
rate
limitations. In the prior art, it was thus generally thought that in order to
support an
XDSL type connection which may contain information at a rate of 128 kbps that,
at
best, only about (500 kbps/128 kbps) or only three (3) subscriber units 101
could be
supported at best on each radio channel.
In contrast to this, the present system 100 subdivides the available radio
channel resources into a relatively large number of subchannels and then
provides a
way to determine how to allocate these subchannels to best transmit data
between
the base station 104 and each of the subscriber units 101, and vice versa. In
the
illustrated example in Fig. 2, the bandwidth is allocated to sixty-four (64)
subchannels 300. It should be understood herein that within a CDMA type
system,
the subchannels 300 are physically implemented by encoding a data transmission
with one of a number of different pseudorandom (PN) or orthogonal channel
codes.
For example, the subchannels 300 may be defined within a single CDMA radio
frequency (RF) carrier by using different orthogonal codes for each defined
subchannel 300. ('fhe subchannels 300 are also referred to as "channels" in
the
following discussion, and the two terms are used interchangeably from this
part
onward).
As mentioned above, the channels 300 are allocated only as needed. For
example, multiple channels 300 are granted during times when a particular
subscriber unit 101 is requesting that large amounts of data be transferred.
In the
preferred embodiment, the single subscn'ber unit 101 may be granted as many as
28
of these channels in order to allow data rates of up to about 5 Mega bits per
second
for an individual subscriber unit 101. These channels 300 are then released
during
times when the subscriber unit 101 is relatively lightly loaded.
Maximum flexibility can be obtained by adjusting coding rates and
modulation types used for each connection, such as the number of channels. One
CA 02615412 2007-11-22
-9-
particular scheme for assigning channel codes, Forward Error Correction (FEC)
code rate, and symbol modulation types is described in U.S. Patent No.
6,973,140,
issued December 6, 2005.
Before discussing how the channels 300 are preferably allocated and
station 104, the base station'104 establishes and allocates a respective data
buffer 211 through 213. Data buffers 211 through 213 store the data that is to
be
transmitted to their respective subscriber units 101. That is, in a preferred
embodiment, there is a separate data buffer in the base station 104 for each
respective subscriber unit 101. As subscriber units enter into and exit out of
communication sessions or connections with base station 104, the number of
buffers
may change. There is always a one-to-one correspondence between the number of
buffers 211 through 213 allocated to the number of subscriber unit 101
communicating with base station 104. The buffers 211 through 213 may be, for
example, queues or other memory structures controlled by software, or may be
hardware controlled fast cache memory.
The particular process which determines how channels are allocated and
deallocated may reside in a data services function disposed within the upper
layers
of the protocols implemented in the base station 104 and subscriber units 101.
Specifically now, referring to Fig. 3, there is shown a protocol layer diagram
such as typically associated with third generation (3G) wireless communication
services. The protocol layers follow the open system interconnect (OSI)
layered
model with a physical layer 120, media access control sub layer 130, link
access
control sub layer 140, and upper communication layers 150. The physical layer
120
provides physical layer of processing such as coding and modulation of the
individual logical channels. Access to the logical channels is controlled by
the
various functions in the MAC sub layer 130 including channel multiplex sub
layer 132, multiplex control channel multiplex sub layer 131, radio link
protocol sub
layer 133, and SRPB 134. The signaling link access control functionality 141
is
provided in the LAC sub layer 140.
Upper layers processing 150 includes upper layer signaling 151, data
services 152, and voice services 153. The particular decision processes to
allocate
CA 02615412 2007-11-22
or deallocate channels to particular network layer connections resides
therefore in a
data services functionality 152 in the upper layers 150. The data services
functionality 152 communicates with the radio link protoco1133 in the MAC sub
layer 130 in order to perform functions such as to send messages to allocate
and
5 deallocate channels from end to end as demand requires.
Turning attention now to Fig. 4, various components of the base station 104
and subscriber units 101 will be described now in greater detail in connection
with
the process for determining when channels should be allocated or deallocated.
Fig. 4 is a more detailed diagram of the implementation of the session
10 oriented buffering scheme implemented in the data services function 152. In
particular, Fig. 4 shows how this is implemented in the base station 104.
Network
layer traffic is routed to the base station 104 using typical network routing
protocols
such as Transmission Control Protocol/Internet Protocol (TCP/IP). At the base
station 104, incoming traffic is separated into individual traffic flows
destined for
separate subscriber units 1-1,101-2, ..., 101-n. The traffic flows may be
separated
such as by examining a destination address field in the TCP/IP header. The
individual traffic flows are delivered first to transport modules 401-1, 401-
2, ...,
401-n with a transport module 401 corresponding to each of the intended
subscnber
units 101. A given transport module 401 is the first step in a chain of
processing
steps that is performed on'the data intended for each subscnber unit 101. This
processing chain includes not only the functionality implemented by the
transport
module 401 but also a number of session queues 410, a session multiplexer 420,
and
transmission buffers 440. The outputs of the various transmission buffers 440-
1,
440-2, ..., 444-n are then assembled by a transmit processor 450 that formats
the
data for transmission over the forward radio links 110.
Returning attention now to the top of the Fig. 4 again, each transport module
401 has the responsibility of either monitoring the traffic flow in such a way
that it
stores data belonging to different transport layer sessions in specific ones
of the
session queues 410 associated with that transport module 401. For example,
transport module 401-1 assigned to handle data intended to be routed to
subscriber
unit 101-1 has associated with it a number, m, of session queues 410-1-1, 410-
1-2,
., 410-1-m. In the preferred embodiment, a given session is characterized by a
particular transport protocol in use. For example, in a session oriented
transport
CA 02615412 2007-11-22
Y1
protocol, a session queue 410 is assigned to each session. Such session
transport
oriented protocols include, for example, Transmission Control Protocol. In
sessionless transport protocols, a session queue 410 is preferably assigned to
each
stream. Such sessionless protocols may for example be the Universal Datagram
Protocol (UDP). Thus traffic destined for a particular subscn'ber unit 101-1
is not
simply routed to the subscriber unit 101-1. First, traffic of different types
are from
the perspective of the transport layer are first routed to individual session
queues
410-1-1, 410-1-2, ..., 410-1-m, associated with that particular connection.
Another key function performed by the transport module 401-1 is to assign
priorities to the individual queues 410-1 associated with it. It will later be
understood that depending upon the bandwidth available to a particular
subscnber
unit 101, traffic of higher priority will be delivered to the transmission
buffer 440-1
before those of lower priority. This may include traffic that is not session
oriented,
for example, real time traffic or streaming protocols that may be carrying
voice
and/or video information.
More particularly, the transport module 401-1 reports the priorities of each
of
the individual session queues 410-1 to its associated session multiplexer 420.
Traffic of higher priority will be selected by the session multiplexer 420 for
loading
into the transmit buffer 440-1 for loading traffic of lower priority, in
general.
Traffic of equal priority will either be fairly selected such as using
techniques know
as weighted fair queuing (WFQ) or other schemes such as oldest queued data
loaded
first.
Priorities associated with each session queue may be obtained from
information such as a profile data record kept for each user. For example,
some
users may have specified that they desire web page traffic traveling on TCP
type
session connections to have lower priority than streaming audio information
carried
on UDP type connections. Prioritization may also be based on other aspects of
the
data content being transmitted. For example, traffic being forwarded from a
private
data network may be given priority over traffic being forwarded from public
networks.
Each of the session multiplexers 420-1, 420-2,..., 420-n, reports indications
to a session manager 430 of the states of all of the session queues 410 that
it is
currently managing. The session manager 430 also receives indications of the
CA 02615412 2007-11-22
f2
present forward channel assignments given to each individual subscriber unit
101 by
the channel assigner 209. The channel assigner 209 monitors the usage of the
transmit buffers 440 in the base station. Upon receipt of characteristic
information
concerning the state of how much data is queued in respect to transmit buffers
440,
the channel resource assigner 209 then determines an urgency factor
representing the
relative need for each subscriber unit 101 to receive data on the available
forward
link radio channels 110. Using these urgency factors, the channel resource
assigner
209 can then dynamically assign an optimum number of channel resources to be
allocated to each subscriber unit 101. Specific discussion of urgency factors
in the
allocation of channels is described in further detail below.
To estimate how much data may be transversing the wired network at any
particular instant in time, the session manager 430 also needs to maintain a
running
estimate of the latency or the back call network 105 to any particular server
at the
other end of a transport layer session. The transport modules 401 therefore
watch
individual session flows from various network servers located in the wired
network
105 and are therefore capable of estimating latencies such as by determining
typical
TCP round-trip time estimations. The transport modules 401 report this
information
to the session manager 430.
The session manager 430 containing all of this information can then send
channel requests to the channel resource assigner 209 when it perceives that
the
present incoming data flow from the wired network for a particular individual
subscn'ber unit 101-1 is greater than the data rate allowed to that subscriber
unit by
its present channel configuration. Recalled from above,that the channel
configuration may include the number of channels assigned, coding rate, and
symbol
modulation rate for each specific channel. Likewise, the session manager 430
notifies the channel resource assigner 209 when it is possible to release
channel
resources for a particular subscriber unit 101-1 if the incoming data flow
from the
wired network 105 is less than the maximum data rate that is presently
assigned to
its forward link.
If split connection transport approaches are employed, (as described in RFC
2757 - Long Thin Networks, see
http://www.ietf,org/rfc/rfc2757.txt?number=2757)
the session manager 430 is capable of sending requests to the transport
modules 401
that pause data flow for any particular session or sessions. If the session is
a TCP
CA 02615412 2007-11-22
13-
session, the transport modules 401 can actively place the TCP senders at the
other end of the network 105 into a s a11ed persist mode, thereby pausing all
$uther session flow. If the session is a g or qnreliable protocol such as
YJDP, a loss profile vvill dete~tain.e the tum of how the queued and incoming
data
is lost. Session knfonnataon will be pa or lost if the session manager 430
requests that more forc~ward bandwidth sj~ould be assigned to a particular
subscariber
unit 101-1 and the request is deuied.
I:f chumei reqttests are denied, e session manager 430 then determines
which session ivformation to regulate, e, or lose data based on content
piiority
infozmation. As previously mentioneci, e tran.sport session rnanagers 401
maintain information to allow them to p'oritize thear individual session
queues
410 based on content so these transport odules 401 can therefore choose the
correct session queues to enable and/or (!isable based on priority.
The transmission bn$'ers 440 each marked with levels that are used to
calculate urgency faators for each resp 've buffer 440. The urgency factors
are
e channel assigrter 209 on a per
used to determine channel allocataon by fl...
ated in Fig, 4 as Ll, L2, and L3,
subscriber per content basis. The levels, c
represent demarcation points for channe allocation andlor deallocatian
SpceificaIly, when tbs transmission buff rs 440-1 are filling and a level is
traversed, an indieation is sent to the c e1 resource assiga,er 209 that the
subscnber utut 101-11s likely to nodt m re forward link bandwidth assigned, If
the request Is denied, the charmel reso assigner 209 then sends this
indication
to the session manager 430.
Conversely, when the transntissi buffer 440-1 is emptying, and a level is
traversed, an indication is sent to the c el resowm assiguer 209 that the
associated subscriber unit 101-1 may h~v forward ttafl'io chattnels taken away
from or deallocated without affecting en to end perform.ance.
The levets L I, L2, L'3, aay there re be texmed under flow tlsresholds.
The levels basically represent pvrmetatio s of available code rate and ehamzel
code
assigncu.ents for an individual subsaxt'ber 't 101. Two requirements are
nceded to
decerrc7ine the ttueshold lavels. F4st, the ute trip trtmsfer time on the
wired
networlC either nee(U to be estnmated or ' tiai approximation needs to be set.
For
TCP sessions, a rmuing rowd-trip time TT) estimation is made, For streaming
orieuted sessions such as Ulap, another roximation can be made which for
CA 02615412 2007-11-22
14
example may be a function of how much data may be queued to optimize the
user's
experience for a particular real time application using the UDP protocol.
Secondly, the data rate over the air interface needs to be determined. This is
a function of the present code rate (CR) and number of assigned channels (NCH)
allocated to a particular subscriber unit. These are the values determined by
the
channel resource assigner 209.
Coding rates are assigned to subscnber units 101 determined by the quality
of the radio connection. For each assigned coding rate, the subscriber may
also be
assigned a number of channels. One scheme, therefore, allocates a Level to
each
available assigned channel. Thus levels L1-LC, where C indicates the number of
assigned channels are available at any given instant in time to service the
connection. Thus the levels, Ll-LC, change each time the number of channels
are
assigned as well as each time the coding rate changes. Specifically, the
particular
buffer level associated with each L will change depending upon the available
coding
rate.
A graphical representation of a particular transmit buffer 440 is illustrated
in
Fig. 5. With knowledge of the round-trip transfer time in the network 105 and
the
current available data rate over the forward link radio channels 110 allocated
to the
particular subscriber unit 101, the levels Ll-LC may be calculated as follows:
Ln = Underflow Threshold = DRA;,.(code rate & channel configuration) * Ot,
where DRair is the data rate across the air interface, and the round-trip
transfer time
is either the estimated time or the set round-trip time over the wired network
105. At
is the time granularity used to monitor incoming data flows. If this scheme is
used
only to optimize TCP connection oriented sessions, At can be said to either
the
maximum or average of all round-trip times estimated by the TCP end points,
depending upon the available buffer space.
The condition for sending a request for more bandwidth to be allocated to a
particular subscn'ber unit 101 is described by the following relationship:
CA 02615412 2007-11-22
[BCa + Fin, * At > L(n + 1)
l r_i J
where At is the time granularity used to monitor the incoming data flows, BCot
represents the current transmission buffer capacity at the beginning of a
particular
5 timeframe, Finl minus Finma, represents all incoming data flows from
sessions or
streams to the transmission buffer 440, and L(n + 1) is the amount of data
that can
be sent over the radio forward links 110 in time m for the next increasing
channel
configuration.
Note that for session oriented TCP streams that the maximum Finsõbi is equal
10 to the maximum advertised received window divided by the round-trip
transfer time.
This condition occurs when the combination of all incoming flows for a
specific
time interval is greater than the amount of data that can be transmitted
during one
time interval At at the next increasing channel capacity assignment.
Fig. 6 represents this case graphically with the block arrow in the Figure
15 representing the amount of flow incoming for the time frame At.
The condition for sending a channel deallocation request for a subscn'ber unit
is given by the relationship:
[BCA, + IY Fin; * At IJ < L(n)
l r=r J
where L(n) is the amount of data that can be sent over the assigned forward
link
channels 110 in time At for the current channel configuration. This condition
occurs
when the combination of all incoming flows for a specific time interval, At is
less
than the amount of data that can be transmitted during that time interval at
the
current channel capacity assignment. This situation is represented in the
diagram of
Fig. 7 with the block arrow representing the amount of flow incoming during
time
At.
Note that in an actual implementation, the transmission buffers 440 may only
be theoretical queues represented by a data structure within the session
manager 430
or session multiplexers 420. The transmission buffers 440 are actually the
combination of all data residing in all session queues 410 for any particular
CA 02615412 2007-11-22
16
subscnber unit 101. This same logic applies when determining urgency factors
and
levels for the transmission buffer data structures namely that such logic can
be
implemented within the session manager 430 and/or session multiplexers 420
rather
than as a separate physical data storage structure and associated logic.
The present invention therefore provides an advantageous way in which
transmission queues may be loaded and how additional resources may be
requested
and/or may be allocated and/or deallocated on a per subscriber basis.
Individual
transmission queues intended for particular subscribers may therefore be
monitored
for data level and channels assigned or deassigned depending upon observed
buffer
filling rates. The channel resource assigner 209 therefore has knowledge of
the
types of traffic flow through the base station based upon application content.
This
allows more intelligent efficient chann,el allocation when there is
competition for the
available resources. Thus by having transport layer aware channel allocation
and
deallocation coupled with calculation of overflow and underflow threshold
based
upon current configured forward link radio channel capacity, the connection
between the base station and tbe subscriber unit in the forward link direction
may be
optimized.
While this invention has been particularly shown and described with
references to preferred embodiments thereof, it will be understood by those
skilled
in the art that various changes in form and details may be made therein
without
departing from the scope of the invention encompassed by the appended claims.
While this invention has been particularly shown and descnbed with
references to preferred embodiments thereof, it will be understood by those
skilled
in the art that various changes in form and details may be made therein
without
departing from the scope of the invention encompassed by the appended claims.
